A method and system for evaluating the performance of a tension belt structure of a sportswear

Abstract

The present application relates to a kind of method and system for evaluating sports suit tension band structure performance, belong to sports clothing performance analysis technical field, solve the problem that evaluation means lacks unified objective standard in prior art, cannot be reused.The method includes: quantitative analysis is carried out to the sports suit to be evaluated, and the tension band is represented using corresponding structure parameter vector;The structure parameter vector of tension band is applied to the parameterized human biomechanics three-dimensional model, and finite element simulation experiment is carried out, and one or more mechanical intermediate variable vectors under the action of tension band are collected;Mechanical intermediate variable vector is used in pre-trained Gaussian process regression proxy model to predict, and the kinematic characteristic change amount before and after the sports suit to be evaluated is obtained;The comprehensive performance of the sports suit to be evaluated is evaluated based on kinematic characteristic change amount.The present application can quantitatively evaluate the performance of sports suit tension band structure, realizes efficient, reusable performance evaluation.

Description

Technical Field

[0001] This invention relates to the field of sportswear performance analysis technology, and in particular to a method and system for evaluating the performance of tension band structures in sportswear. Background Technology

[0002] In competitive sports such as running, bobsleigh, speed skating, and weightlifting, or in sports activities such as dance, gymnastics, and fitness, the performance of sports equipment, especially the design of tension band structures, has an increasingly significant impact on sports assistance, individual fit, and muscle / joint protection. More and more sports equipment needs to be designed and optimized from a more scientific perspective in order to provide more reasonable and targeted assistance and protection for equipment users.

[0003] Taking the design and optimization of tension band structures in sportswear as an example, the fit and mechanical performance of the sportswear after wear has become a key evaluation item. Currently, the performance evaluation of tension band structures in sportswear is usually conducted through laboratory physiological and biomechanical tests, questionnaires and performance feedback, or posture evaluation based on machine vision. These methods either rely on customized experiments, subjective feedback, or post-hoc analysis of the effects of sportswear based on individual postures and movements (i.e., customized sportswear is developed first and then analyzed one by one).

[0004] It is evident that existing analytical and evaluation methods lack unified standards. Differences in experimental conditions and personnel often lead to significant variations in results, limiting their reference value. Furthermore, existing technologies lack quantitative correlation analysis methods. The analysis process can only be conducted after the sportswear design is completed, and each analysis is conducted independently for specific sportswear and personnel. The analysis and evaluation process and conclusions for each sportswear are isolated from each other, making it difficult to share evaluation experiences and results and to dynamically adjust and provide feedback. This results in low efficiency and non-reusability. Summary of the Invention

[0005] Based on the above analysis, the embodiments of the present invention aim to provide a method and system for evaluating the structural performance of tension bands in sportswear, in order to solve the problem that the evaluation methods in the prior art lack unified objective standards and cannot be reused.

[0006] On one hand, embodiments of the present invention provide a method for evaluating the structural performance of tension bands in sportswear, the method comprising the steps of:

[0007] A quantitative analysis is performed on the sportswear to be evaluated, and one or more tension bands in the sportswear to be evaluated are represented by one or more corresponding structural parameter vectors.

[0008] The structural parameter vectors of the one or more tension bands are applied to the parameterized human biomechanical three-dimensional model, and finite element simulation experiments are performed through the parameterized human biomechanical three-dimensional model to collect the mechanical mediator variable vectors under the action of the one or more tension bands.

[0009] The changes in kinematic characteristics before and after wearing the sportswear to be evaluated are obtained by using the aforementioned mechanical mediating variable vector in a pre-trained Gaussian process regression surrogate model.

[0010] The overall performance of the sportswear to be evaluated is assessed based on the changes in the kinematic characteristics.

[0011] Based on a further improvement of the above method, the structural parameter vector includes at least the tension direction, effective area, and pretension of the tension band. Applying the structural parameter vectors of the one or more tension bands to the parameterized three-dimensional human biomechanical model includes the following steps:

[0012] Each of the tension bands is modeled as a membrane unit in the parametric human biomechanical 3D model;

[0013] The tension direction is reflected by the local material orientation of the membrane unit or the direction of the applied load;

[0014] The effective working area is parametrically controlled by adjusting the thickness of the membrane unit and the covered grid area.

[0015] The prestress is applied by setting the initial stress of the membrane unit in a predefined field.

[0016] Based on a further improvement of the above method, the structural parameter vector also includes the tension band action area, and the mechanical mediating variable vector includes: the maximum stress of the target muscle in the tension band action area, the maximum strain of the target muscle, the relevant joint reaction force, and the interfacial shear stress between the tension band and the skin.

[0017] Based on further improvements to the above method, the kinematic features include: joint angles of at least one related joint, joint angular velocities of at least one related joint, trunk stability indices, and symmetry indices; wherein, the at least one related joint is determined based on the tension band action area.

[0018] Based on a further improvement of the above method, the Gaussian process regression surrogate model adopts the Matern 5 / 2 kernel function as the covariance function;

[0019] The pre-training optimizes the hyperparameters and noise level of the kernel function through maximum likelihood estimation, and evaluates the predictive performance of the surrogate model through leave-one-out cross-validation, so that the conditional distribution of the kinematic feature changes predicted by the Gaussian process regression surrogate model based on the mechanical mediating variable vector is a Gaussian distribution.

[0020] Based on a further improvement of the above method, the kernel function is:

[0021] ;

[0022] Among them, the independent variable , and It is the input vector of any two sets of samples. It is the signal variance. It is a length scale parameter.

[0023] Based on a further improvement of the above method, the maximum likelihood estimation obtains the optimal hyperparameters of the kernel function by maximizing the log-marginal likelihood function, which is:

[0024] ;

[0025] in For the kernel matrix, It is the identity matrix. , These are the input matrix and the output matrix, respectively. This is the transpose of the output matrix; the dimensions of each matrix are related to the number of samples. The hyperparameters of the kernel function are... denoted as the noise level of the kernel function, and n is the total number of training samples.

[0026] Based on further improvements to the above method, the leave-one-out cross-validation method calculates the root mean square error (RMSE) and the coefficient of determination. As an evaluation indicator:

[0027] ;

[0028] ;

[0029] in, These represent the i-th true label value, the i-th model predicted output value, and the mean of the true labels, respectively; N is the total number of predicted values.

[0030] Based on a further improvement of the above method, the overall performance of the sportswear to be evaluated is assessed by calculating a comprehensive utility score U:

[0031] ;

[0032] in, Let be the directional change of the k-th kinematic feature; Let be the reference change for the k-th kinematic feature. Let k be the weight coefficient of the k-th kinematic feature under a specific motion type, satisfying d represents the total number of kinematic characteristics. It is the absolute change of the k-th kinematic feature directly predicted by the model. It is the coefficient of the benefit direction sign.

[0033] On the other hand, embodiments of the present invention provide a system for evaluating the structural performance of tension bands in sportswear, the system comprising:

[0034] The tension band quantification module is used to perform quantitative analysis on the sportswear to be evaluated, and to represent one or more tension bands in the sportswear to be evaluated using one or more corresponding structural parameter vectors.

[0035] The finite element simulation module is used to apply the structural parameter vectors of one or more tension bands to a parameterized human biomechanical three-dimensional model, and to conduct finite element simulation experiments through the parameterized human biomechanical three-dimensional model to collect the mechanical mediator variable vectors under the action of one or more tension bands.

[0036] The surrogate model module is used to make predictions in a pre-trained Gaussian process regression surrogate model using the mechanical mediating variable vector to obtain the changes in kinematic characteristics before and after wearing the sportswear to be evaluated.

[0037] The performance evaluation module is used to evaluate the overall performance of the sportswear to be evaluated based on the changes in the kinematic characteristics.

[0038] Compared with existing technologies, the method for evaluating the performance of tension band structures in sportswear proposed in this invention significantly outperforms existing technologies in several aspects by constructing an evaluation method system that integrates tension band structure and human posture characteristics:

[0039] 1. A unified and standardized tension band design description system

[0040] This invention introduces a set of design parameters to describe the force, area, direction, angle, and region of action of each tension band, and establishes a standardized numbering library, enabling quantitative comparison of the structures of different sports and types of clothing within a unified framework. This method solves the problems of "experience-based design, inconsistent descriptions, and lack of data interoperability" in traditional sportswear, supporting product database construction and large-scale statistical analysis. It allows different design teams to share a unified design language, avoiding errors caused by inconsistent understanding of terminology or structures, and shortening cross-team collaboration cycles.

[0041] 2. Enables a label-free, natural evaluation process with wide applicability.

[0042] This invention utilizes a markerless posture recognition algorithm combined with an interactive correction mechanism to achieve accurate identification of key joints and movement features of athletes without relying on optical motion capture, inertial sensors, or human body marker attachments. The method is simple to operate, the evaluation process does not interfere with natural movement, and it is suitable for complex scenarios such as real training, outdoor testing, and competition analysis. It overcomes the limitations of laboratory reliance and high costs, reducing costs by more than 90% (requiring only a regular camera vs. a megapixel-level optical capture system), thus improving the practicality and scalability of the evaluation technology.

[0043] 3. Construct a structure-action fusion model to achieve interpretability analysis of clothing utility.

[0044] Using a technical approach of "parametric finite element simulation + Gaussian process regression surrogate model", a model was constructed based on the tension band design parameters (F, A, ...). This study establishes a quantitative and interpretable mapping relationship between tension band structures and biomechanical mediators (muscle stress, joint forces), and then to changes in kinematic characteristics. It clarifies the positive or negative contribution of tension band structures to specific movements, thereby enabling interpretable and quantifiable evaluation of the utility of functional clothing and providing a scientific basis for design optimization.

[0045] 4. Applicable to various action scenarios and professional sports, with strong versatility.

[0046] This invention possesses excellent versatility in its algorithm and modeling methods, applicable not only to the evaluation of basic movements (such as running, squatting, and pushing a cart), but also to competitive sports that demand higher levels of postural stability and explosive movement, such as bobsleigh, speed skating, and weightlifting. By appropriately configuring movement templates and analysis parameters, it can be flexibly extended to the adaptability evaluation of different sports and functional clothing.

[0047] 5. Low equipment requirements, controllable implementation costs, and easy operation.

[0048] Compared to traditional physiological assessment platforms or high-end motion capture systems, this invention only requires conventional shooting equipment (such as high-definition cameras, action cameras, or mobile phones) and structural parameter input (CAD drawings, image annotations, etc.). It does not require additional high-cost hardware equipment and is not subject to space constraints, providing a low-threshold and usable solution for small and medium-sized laboratories, corporate clothing testing, and grassroots sports units.

[0049] In this invention, the above-described technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of this invention will be set forth in the following description, and some advantages may become apparent from the description or be learned by practicing the invention. The objects and other advantages of this invention can be realized and obtained from what is particularly pointed out in the description and drawings. Attached Figure Description

[0050] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts.

[0051] Figure 1 This is a schematic flowchart of a method for evaluating the structural performance of tension bands in sportswear according to one embodiment of the present invention;

[0052] Figure 2 This is a schematic diagram illustrating the human body key points provided by a markerless pose recognition algorithm in one embodiment of the present invention. Detailed Implementation

[0053] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0054] Currently, the performance evaluation of tension band structures in sportswear is typically conducted through laboratory physiological and biomechanical tests, questionnaires and performance feedback, or posture assessment based on machine vision. Laboratory testing methods generally rely on biomechanical experimental platforms, using high-end equipment such as electromyography (EMG), force platforms, and 3D optical motion capture systems to precisely collect and analyze human motion characteristics before and after wearing sportswear in a controlled environment. While this method offers high accuracy and detailed data, it has the following limitations: expensive equipment, complex operation, and high experimental costs, making it unsuitable for widespread adoption; the operation process interferes with the subjects, affecting their natural movement and causing some distortion; it can only be conducted in a laboratory setting, which is spatially limited and makes it difficult to simulate real training or competition environments; and it requires attaching markers to the body, the position of which is easily affected by clothing obstruction and slippage, causing data errors, making it particularly unsuitable for evaluating the structure of tight-fitting clothing.

[0055] The effectiveness of clothing is typically assessed indirectly through questionnaires, feedback on wearing experiences, and changes in performance before and after use (such as running speed and weight gain). While these methods are simple and inexpensive, they also have significant drawbacks: they lack objective quantitative indicators, making it difficult to systematically analyze the coupling relationship between the tension band structure of sportswear and body posture; the results are strongly correlated with individual circumstances and subjective feelings, easily influenced by factors such as the subject's subjective perception ability, experience level, and psychological state, lacking repeatability and persuasiveness; and they cannot provide structural-functional explanations, hindering design optimization and scientific improvement.

[0056] Machine vision research is mostly used in scenarios such as motion monitoring, rehabilitation assessment, and sports training. Research on the performance of tension band structures in sportswear is still in its early stages and has the following shortcomings: existing posture recognition methods are mostly trained based on images of naked human bodies, and the recognition accuracy decreases when wearing tight clothing or obscuring the body contour; there is a lack of correlation modeling with clothing structural parameters (such as tension band position and direction), making it difficult to clarify the mechanism of clothing action; there is a lack of a structure-motion coupling analysis framework, making it impossible to identify the promoting or inhibiting effects of functional clothing on key motion characteristics.

[0057] Therefore, there is currently no unified and systematic method to achieve interpretable modeling and quantitative evaluation of changes in human movement before and after wearing without marking. This is especially true for functional clothing with active guidance structures (such as tension bands), where the mechanism of action and the relationship between action response are more complex. Existing technologies cannot meet the needs of functional verification, design optimization, and personalized recommendations for this type of product.

[0058] To address the aforementioned problems in existing technologies, this invention provides a method for evaluating the performance and utility of functional sportswear that integrates tension band structural parameters and human posture characteristics, applicable to label-free video data analysis. It is particularly suitable for evaluating the motion assistance, posture control, and support effects of tension band structure sportswear, solving problems such as insufficient accuracy, limited application scenarios, and difficulty in quantitative analysis in existing evaluation methods.

[0059] One specific embodiment of the present invention discloses a method for evaluating the structural performance of tension bands in sportswear, such as... Figure 1 As shown, the method includes the following steps:

[0060] S11, Perform quantitative analysis on the sportswear to be evaluated, and represent one or more tension bands in the sportswear to be evaluated using one or more corresponding structural parameter vectors;

[0061] S12, apply the structural parameter vectors of the one or more tension bands to the parameterized human biomechanical three-dimensional model, and conduct finite element simulation experiments through the parameterized human biomechanical three-dimensional model to collect the mechanical mediator variable vectors under the action of the one or more tension bands;

[0062] S13, use the mechanical mediating variable vector to make predictions in a pre-trained Gaussian process regression surrogate model to obtain the changes in kinematic characteristics before and after wearing the sportswear to be evaluated.

[0063] S14, evaluate the overall performance of the sportswear to be evaluated based on the changes in the kinematic characteristics.

[0064] Existing posture recognition-based analysis methods generally neglect the structural features of sportswear, failing to reflect the intervention effect of tension bands, a typical structure, on target joints or muscle groups in the human body. This invention, through a quantitative and parametric description of the tension band structure and coupled modeling with posture features, achieves a quantitative correlation between the structural mechanism and athletic performance, helping to explain the specific impact of clothing structure on athletic characteristics.

[0065] In this embodiment of the invention, by decomposing the function of the tension band structure, it is determined that its assistance to the function of human joints / muscle groups depends on parameters such as direction, area, strength, and area of ​​action. Therefore, the tension band related information is extracted by 3D scanning model to construct the corresponding tension band structure parameter vector.

[0066] Specifically, each sportswear garment may provide multiple tension bands to assist movement in different areas. To achieve a unified utility assessment of functional clothing for different types of sports, the tension band structure is parametrically coded and uniformly modeled. All tension bands available in the sportswear are numbered to construct a tension band set. C represents the total number of tension bands. Considering the differences in the number, location, and direction of tension bands in clothing for different sports, the following principles are used to construct a complete set of tension bands shared by different types of sportswear:

[0067] If the tension bands have identical structures (same origin, direction, and area of ​​action), they are numbered the same across garments. If the direction of action, anchor point location, or coverage differs, different numbers are assigned even if they act on the same anatomical region. This ensures that different types of sportswear can be uniformly processed under the same mathematical model, and specific sportswear evaluation and analysis only requires selecting the corresponding subset of tension bands.

[0068] For the i-th tension band The structural parameters of the tension band are quantified using a standard 3D human body model to construct the structural parameter vector of the tension band. ,in, The tension direction vector is represented as a unit vector in a three-dimensional coordinate system. The origin of this system is located at the center of the human pelvis; the Y-axis is the vertical axis, pointing directly upwards; the X-axis is the horizontal axis, pointing horizontally to the right; and the Z-axis is the sagittal axis, pointing horizontally backwards. This coordinate system conforms to the anatomical coordinate system commonly used in biomechanical analysis, is fixed to the human body, and does not change with limb movement. The actual spatial direction of the tension band is measured on a clothing CAD model or a standard human body model. This direction is typically represented by a directed line segment, pointing from the proximal attachment point of the tension band to the distal attachment point. After measuring the actual direction vector, it is mathematically normalized, and the "unit 1" is obtained by dividing by the vector magnitude. This represents the effective area of ​​the tension band. The expected unit force or pretension intensity to be applied (e.g., derived from physical simulation or actual measurement); The target area of ​​the tension band (such as the hip joint, knee joint, shoulder muscles, etc., and there may be multiple areas) is represented by standard human body part codes in this embodiment of the invention. The specific codes and their correspondence with major human muscle parts are detailed in Table 1 below:

[0069] Table 1

[0070]

[0071] Typically, running apparel enhances lower limb and trunk stability. The structural parameter vectors of several tension bands commonly used in running apparel can be described as follows:

[0072] [R10 = gluteus medius, R16 = biceps femoris]};

[0073] [R12=rectus femoris, R14=vasomotor]};

[0074] , [R19=gastrocnemius, R20=soleus]}.

[0075] Skating apparel emphasizes trunk stability and lower limb abduction control. The structural parameter vectors of several tension bands commonly used in skating apparel can be described as follows:

[0076] , [R10=gluteus medius, R09=gluteus maximus]};

[0077] , [R11=tensor fasciae latae]};

[0078] , [R21=Tibialis anterior]}.

[0079] In this way, all tension bands that can be used in sportswear are described and managed in a unified quantitative manner in this embodiment of the invention, thereby laying the foundation for digital processing and model interpretability.

[0080] More specifically, this embodiment of the invention also constructs a parameterized three-dimensional human biomechanical model to achieve precise quantitative analysis of the biomechanical impact of the tension band structure of sportswear on the human body. Preferably, a parameterized human biomechanics-clothing coupling model is established in Abaqus / CAE, and the mechanical state variables of the target area are obtained through simulation.

[0081] First, geometric reconstruction and mesh generation are performed using the collected human body data to obtain a basic 3D human body model. For example, based on medical imaging data (CT / MRI) of a healthy male volunteer with a height of 175cm and a weight of 70kg, 3D reconstruction is performed using Mimics 21.0 software to obtain geometric models of the pelvis, femur, tibia and fibula, foot bones, and major muscles and soft tissues such as the gluteus maximus and biceps femoris. The model is imported into 3-matic in STL format for smoothing and solidification to generate a volume mesh model. Finally, the processed geometric model is imported into Abaqus for mesh generation. Tetrahedral elements (C3D4) are used for bones, and hexahedral-dominated hybrid elements (C3D8H) are used for muscles and soft tissues to ensure computational stability under large deformations.

[0082] The mechanical properties of the basic 3D human body model are then constrained. Typical constraints include material property assignment and boundary load condition setting. Material property assignment includes:

[0083] Cortical bone: considered as an isotropic linear elastic material, with an elastic modulus E = 17 GPa and a Poisson's ratio ν = 0.3;

[0084] Cancellous bone: considered as an isotropic linear elastic material, with an elastic modulus E = 1 GPa and a Poisson's ratio ν = 0.2;

[0085] Muscle: The nonlinear mechanical behavior was simulated using the Ogden hyperelastic model (model order 3). The material parameters were: μ1=0.0185 MPa, μ2=0.0099 MPa, μ3=0.0032 MPa, α1=22.39, α2=14.71, α3=6.42; the quasi-static incompressibility assumption was D1=D2=D3=0.

[0086] Fat / subcutaneous soft tissue: The Ogden hyperelastic model (model order 2) was used with the following parameters: μ1=0.0021 MPa, μ2=0.0015 MPa; α1=10.00, α2=-7.50; D1=D2=0;

[0087] Skin: A linear elastic material model was adopted, with an elastic modulus E = 4.5 MPa and a Poisson's ratio ν = 0.45;

[0088] Clothing fabric (tension band substrate): simulated using an isotropic linear elastic shell element (S4R), with an elastic modulus E=120MPa and Poisson's ratio ν=0.4.

[0089] Boundary load condition settings include:

[0090] Fix all degrees of freedom on the upper surface of the pelvis to simulate the mid-stance of standing while running;

[0091] Muscle forces optimized based on OpenSim simulation were applied to the distal femur and proximal tibia.

[0092] The measured ground reaction force (GRF) was applied to the sole of the foot, with a peak force of 2.5 times the body weight.

[0093] Then, the tension band structures of the sportswear are parametrically modeled and applied to the 3D human body model. The tension bands are modeled as membrane units (M3D4R), whose material is defined as isotropic linear elastic, but the pretension F is applied through the initial stress in the predefined field.

[0094] The effective working area A of the tension band is parametrically controlled by adjusting the membrane thickness and the covered grid area.

[0095] direction of tension band and the direction cosine calculated therefrom This is reflected through local material orientation or the direction of applied load. These properties can be modified parametrically via the Abaqus scripting interface (Python).

[0096] To ensure the parameterized tension band structure functions realistically and effectively on the three-dimensional human body model, this embodiment of the invention also employs a central composite design (CCD) to arrange simulation experiments, using the three key parameters of the tension band as design variables, and using finite element analysis (FEA) to extract simulation results.

[0097] The three key parameters of the tension band include: pretension F, ranging from 8N ​​to 16N; effective area A, ranging from 40cm² to 60cm²; and direction cosine. The variation range is 0.70-0.95.

[0098] In the Design of Experiments (DOE), 20 parameter combinations are generated based on CCD (including 8 factorial points, 6 pivot points, and 6 center points): the 8 factorial points represent "corners" in the design space and are part of the 23-factor design, with a coding level of (±1, ±1, ±1), for example: (F:-1, A:-1, :-1), (F:-1, A:-1, (F:+1), ..., (F:+1, A:+1, +1); Six axis points (star points) are located on the coordinate axes of each factor and are used to estimate curvature. Their coding levels are (±α, 0, 0), (0, ±α, 0), and (0, 0, ±α), for a total of six points, where α = 1.68179; six center points are points where all factor levels are 0 (0, 0, 0), repeated six times, used to estimate experimental error and model stability. The "coding level" of a factor characterizes the proportion of the actual parameter value's deviation from the design center point. Specifically, a coding level of ±1 indicates that the change in the factor value relative to the center point is equal to half the range of the parameter's change, i.e., half the parameter's variation is represented by a unit of 1; while the coding level ±α (where α = 1.68179) corresponds to the axis point position of each factor in the design space, reflecting the secondary effect and curvature characteristics of the response surface. The selection of the α value ensures the rotatability of the design, i.e., the model's prediction variance is uniform in all directions.

[0099] All simulation tasks were submitted to the high-performance computing (HPC) cluster for solving using Abaqus scripts in batches. The average computation time for each model was approximately 4 hours (using 128 CPU cores).

[0100] After each simulation calculation is completed, the following vector of mechanical mediating variables is automatically extracted using a Python script:

[0101] Mechanical response of target muscle / soft tissue: Extracting the target area of ​​action (from tension band parameters) (Definition) Maximum principal stress of muscle or soft tissue (Unit: MPa) and maximum principal strain It is used to assess the load and deformation of an organization;

[0102] Resultant contact forces of relevant joints: Extracting the joint reaction force (JRF) of the major joints related to the biomechanical function of the target action area. (Unit: N). For example, if the tension band acts on the gluteus maximus (R09) or gluteus medius (R10), the hip joint reaction force is extracted; if it acts on the rectus femoris (R12) or biceps femoris (R16), the hip and knee joint reaction forces can be considered simultaneously; if it acts on the gastrocnemius (R19), the ankle joint reaction force is extracted. The extracted reaction forces of the major joints reflect the total load on the joint surfaces.

[0103] Interfacial mechanical properties: Extracting the interfacial shear stress between clothing and skin (Unit: MPa) is used to evaluate the fit and comfort of clothing.

[0104] Taking a compression shorts designed to improve running performance as an example, its two core tension band parameters are as follows:

[0105] : Acts on the gluteus maximus (R09), with the following design parameters. ;

[0106] : Acts on the biceps femoris (R16), with the following design parameters. .

[0107] Table 2 below shows two sets of simulation results that are similar to the design points mentioned above in the CCD experimental design:

[0108] Table 2

[0109]

[0110] Simulation results show that tension band and The introduction of this effectively increases the stress on the target muscle, meaning the muscle is activated in response to external assistance, while quantifiable changes occur in the hip joint contact force. These mechanical mediating variables (muscle stress) ,strain Joint force Interfacial shear stress This provides reliable input data for the next step of building a surrogate model to predict athletic performance.

[0111] After quantifying the structure of each tension band, this embodiment of the invention further extracts the kinematic features corresponding to each sample sportswear. In this embodiment, firstly, based on the key point data obtained from video pose recognition, multiple kinematic features and their changes, such as joint angles, joint angular velocities, stability indices (trunk offset), and symmetry indices, are calculated. Specifically, a camera is used to capture motion videos of the subject before and after wearing the sample sportswear (preferably, the motion videos can be motion simulation videos of the three-dimensional human body model established in the aforementioned simulation experiment), obtaining motion image data before and after wearing; subsequently, a markerless pose recognition algorithm (such as MediaPipe) is used to extract multiple human key points to form sequences from the motion image data before and after wearing. ;in Let be the three-dimensional coordinates of the i-th human body key point at time t.

[0112] The pre-wear sequence can be extracted only after the initial acquisition of motion images before wearing the garment, and the pre-wear sequence J0 can be used uniformly thereafter, without having to re-acquire and extract for each sample of sportswear.

[0113] like Figure 2 As shown, MediaPipe marks 33 typical key points on the human body, and the specific correspondences can be seen in Table 3 below. Some of these points are not points that sportswear needs to cover and control. For example, the 11 points from 0 to 10 are head points, and the 6 points from 17 to 22 are hand detail points. Tension band structures are generally not set in these areas. In this embodiment of the invention, the tension band structure of sportswear is not considered for the time being when evaluating it, and it can be skipped during identification.

[0114] Table 3

[0115]

[0116] When wearing tight-fitting functional clothing, traditional posture recognition algorithms are prone to problems such as misidentification, drift, or loss of key points due to factors such as fabric obstruction and tension deformation. This invention employs an interactive recognition optimization mechanism, which first automatically identifies outliers in key points through algorithms, and then corrects erroneous key points through manual interaction. This effectively improves the stability and accuracy of posture data, ensuring the effective extraction and analysis of subsequent kinematic features.

[0117] The first step involves detecting anomalous key points by setting physical constraint rules. In pose recognition, common anomalies are mainly joint displacement or joint angle anomalies. The specific constraints for detecting these two types of anomalies include spatial continuity constraints and anatomical angle constraints. The former limits the threshold of joint displacement between adjacent frames, while the latter limits the range of joint angles identified in each frame.

[0118] Preferably, the joint displacement threshold between adjacent frames is determined by deriving the relational formula. ; Δt represents the maximum joint movement velocity of the human body, and Δt represents the time interval between adjacent frames. The spatial continuity constraints in a preferred embodiment of the present invention are set as shown in Table 4 below:

[0119] Table 4

[0120]

[0121] Anatomical angle constraints limit the range of joint angles to During pose recognition, if any human key point data is found to exceed the range of the two types of constraints mentioned above, it is marked as an anomaly, thus obtaining a set of all anomalies during the recognition process. .

[0122] The system then automatically marks the abnormal frames and error key points based on the set of abnormal points, and then adjusts the errors and restores the abnormalities through automatic or manual tools (typically, the error key points can be dragged to the correct dissection position).

[0123] Furthermore, the corrected data is input into a spatiotemporal smoothing filter:

[0124] ;in, This is the sequence of key points extracted at time t. The sequence of key points at time i is given; the preferred weight is λ=0.7, and the preferred sliding window size is w=5; thus, the optimized key point sequence is obtained. .

[0125] Based on keypoint data obtained from video pose recognition (preferably using outlier detection and correction), multiple kinematic features are calculated, including joint angles, joint angular velocities, stability indices (torso offset), and symmetry indices. Specifically, based on the coordinates of multiple keypoints related to the joint keypoint jm before and after it, the vector dot product formula is used to calculate the current angle of joint keypoint jm (i.e., the current joint angle of joint jm). ).by Figure 2 For example, the current angle of the right hip joint J24. It was obtained using several key points on the same side: j12 (right shoulder joint), j24 (right hip joint), and j26 (right knee joint); the current angle of the right knee joint j26. It is obtained using several key points on the same side: j24 (right hip joint), j26 (right knee joint), and j28 (right ankle joint); the current angle of the right ankle joint j28. It was obtained by using several key points on the same side: j26 (right knee joint), j28 (right ankle joint), and j32 (right toe).

[0126] The value of a vector dot product is the sum of the products of the coordinates of two vectors. It represents the product of the magnitudes of the two vectors and the cosine of the angle between them, and thus the corresponding angle can be calculated. Taking the right hip joint j24 as an example, its coordinates ( The right shoulder joint j12 and coordinates ( The vector formed by the right hip joint j24) and coordinates ( The right hip joint j24 and coordinates ( The vector formed by the right knee joint j26. With common constraints, the relevant vector dot product is:

[0127] ;

[0128] The magnitudes of the two vectors are respectively:

[0129] ;

[0130] ;

[0131] Therefore, the current angle of the right hip joint j24 can be calculated. for:

[0132] .

[0133] Joint angular velocity is the change in angle of a joint per unit time. It can be calculated by dividing the angle difference by the time difference between the two frames of a video (any two frames during a joint movement, but usually two adjacent frames are selected, or the two frames at the beginning and end of the joint movement can be selected). Taking the right hip joint j24 as an example, there is a joint angle difference between the two frames before and after the current time t. If the time interval between two adjacent frames in the video is Δt, then the angular velocity at the current time t is:

[0134] .

[0135] Based on the target area of ​​the tension band, the tension band structural parameter vector can be associated with one or more kinematic features, such as the previous tension band. Extending from the gluteus medius (coded R10) to the biceps femoris (coded R16), the target area corresponds to the hip joint, thus creating a tension band. The structural parameter vector and the angle of the right hip joint j24 and angular velocity Correlation. Based on this method, the correspondence between the tension band target area and anatomical joints (key points in the human body) is shown in Table 5 below:

[0136] Table 5

[0137]

[0138] Furthermore, the kinematic features in this embodiment of the invention may also include stability indices (torso offset) and symmetry indices. In this embodiment, the torso offset, represented by the change in the center of mass position, is used as the stability indices. The default coordinate system is a camera coordinate system (right-handed), where the x-axis is horizontal (left negative, right positive), the y-axis is vertical (top negative, bottom positive), and the z-axis points towards the camera (front negative, back positive). The current center of mass (COM) coordinates are... The value is obtained by averaging the coordinates of two points j11 and j12 on the shoulder and two points j23 and j24 on the hip joint. The vertical fluctuation is mainly reflected on the vertical axis.

[0139] ,

[0140] Determine the degree of vertical fluctuation of the center of mass over time, and describe the stability of the torso based on the change in the position of the center of mass:

[0141] Trunk stability index (trunk offset) ;

[0142] In the formula, Total time used This represents the y-coordinate of the centroid at time t. This represents the average y-coordinate over all times, i.e.:

[0143] .

[0144] Preferably, embodiments of the present invention further use the quantification of the difference in stride length between the left and right legs as a symmetry indicator. This is first measured by the vertical velocity at the key ankle point. Identify the moment of landing:

[0145] Landing time When the vertical acceleration exceeds the threshold γ, it is determined to be a landing.

[0146] Calculate the stride length for each leg separately, defining the stride length as the horizontal displacement of the center of mass (the displacement of COM in the x-direction) between two consecutive landings:

[0147] ,

[0148] COM(t) represents the coordinates of the human body's center of mass at time t (as mentioned above, it is calculated from the average coordinates of the hip and shoulder key points). The x-axis represents the projected distance in the direction of movement; i and j represent the gait cycle indices of the left and right legs. Let i be the time corresponding to the i-th cycle of the left leg. This corresponds to the time of the j-th cycle of the right leg. and These are the calculated stride lengths for the left and right legs, respectively.

[0149] To further quantify the difference in stride length between the left and right legs, there is a symmetry index formula:

[0150] ,

[0151] in, Indicates the number of effective gait cycles; This represents the step length of the left and right legs within the i-th cycle.

[0152] In this embodiment of the invention, the human kinematic characteristics used to evaluate the performance of the tension band structure of sportswear may include: joint angle θ, joint angular velocity ω, and trunk stability index (trunk offset). Symmetry index The evaluation of the performance of the tension band structure in sportswear mainly involves assessing the changes in the kinematic characteristics of the human body after wearing a particular sportswear garment, and judging the quality of the tension band structure performance based on the positive and negative effects caused by these changes.

[0153] The differences in kinematic characteristics before and after wearing the garment are expressed quantitatively:

[0154] ;

[0155] in These are the feature vector values ​​before wearing. These are the feature vector values ​​after wearing the garment. The difference between the two can be identified by the joint angle θ, joint angular velocity ω, and trunk stability index (trunk offset) that constitute vector M. Symmetry index The vector (ΔM) formed by the differences between these feature dimensions is used to represent the feature.

[0156] To achieve accurate and efficient mapping from the mechanical mediating variable vectors (muscle stress, joint reaction force, etc.) obtained from finite element simulation to the changes in macroscopic kinematic characteristics, this invention introduces the Surrogate Model technique to construct a "mechanical response-motion performance" prediction model, thereby avoiding time-consuming physical simulation and enabling rapid evaluation of clothing utility.

[0157] Input variable (X): A vector of mechanical mediating variables extracted from each finite element simulation result. For clothing acting on m target regions, its input characteristics can be represented as:

[0158] ;

[0159] in, , These are the maximum principal stress and strain of the i-th target muscle. It is the main joint reaction force associated with the i-th target. It is the shear stress at the interface corresponding to the i-th target.

[0160] The sample label (i.e., the model's predicted output) is the change in kinematic features ΔM identified before and after wearing the sample sportswear. The change in kinematic features is correlated with the aforementioned mechanical mediating variable vector to construct the sample data. Specifically, for the sample sportswear, 20 finite element simulations are performed based on a central composite design (CCD). Data from c motion cycles is collected for each simulation result, ultimately forming a sample dataset with a sample size of 20 × c. .

[0161] Specifically, in the embodiments of the present invention, the tension band structure and target action area included in each type of sportswear are determined, and the corresponding affected muscles and joints are also determined. Therefore, a model can be constructed to perform a unified quantitative analysis of the tension band structure of this type of sportswear. For example, a certain type of sportswear typically includes *a* tension band structures, quantified into *a* structural parameter vectors (each vector is a triplet of tension direction, effective action area, and pretension force); these *a* tension band structures act on *m* target areas. Simulation experiments are conducted in a parameterized three-dimensional human biomechanical model, corresponding to the extraction of *m* mechanical mediating variable vectors (each vector is a quadruplet of target muscle maximum stress, target muscle maximum strain, related joint reaction force, and interface shear stress); a finite element simulation model is established based on a central composite design CCD, and 20 simulation experiments are conducted according to the 20 parameter combinations of the CCD (8 factorial points, 6 axis points, and 6 center points). The *m* mechanical mediating variable vectors extracted in each simulation are used as input values ​​for a prediction model. 20 simulations extract 20 input values. .

[0162] Then, perform posture recognition on the motion videos before and after wearing the sportswear for multiple motion cycles. Each motion cycle can obtain a set of kinematic feature changes (each set may include 1 stability index, 1 symmetry index, b joint angles—determined by m target regions, and b joint angular velocities—determined by m target regions; the specific kinematic features selected can be reduced by removing unimportant parts according to the actual impact). This set of kinematic feature changes is used as the corresponding output label (true label). .

[0163] By using the mechanical mediator variables obtained through simulation Compared with the measured kinematic change Pairing was performed to form a sample dataset (c groups of measured data, sample size 20×c), resulting in 20 different combinations of tension band design parameters based on the central composite design (CCD). In physical simulations or experimental verification, for each specific parameter combination ( It is necessary to collect data from the subject over c complete movement cycles to obtain stable and statistically significant output. Therefore, the final dataset contains 20 (design parameters) × c (motion cycles) = 20 × c samples. A machine learning model is trained to establish an X-Y mapping relationship, enabling the construction of a surrogate model that predicts the change in kinematic characteristics Y using the mechanical mediator variable X. Subsequently, for a new design of this type of sportswear (the sportswear to be evaluated), which has a new structural parameters (only some parameters may be modified), a simulation experiment is run to obtain a new mechanical mediator variable vector X. The pre-trained machine learning model then predicts the corresponding new change in kinematic characteristics Y. Finally, this embodiment of the invention quantitatively analyzes the positive and negative effects of the new design through the new change in kinematic characteristics Y, thereby evaluating the structural performance of the new design.

[0164] Given the potentially complex nonlinear characteristics of the "mechanical response-motion performance" relationship, this invention preferably employs Gaussian Process Regression (GPR) as the core surrogate modeling method. GPR is a nonparametric Bayesian model that does not use a learnable, fixed-parameter function form but directly defines the distribution of the function's output, making it particularly suitable for handling small-sample, high-dimensional nonlinear problems. GPR not only provides predicted values ​​but also estimates of the uncertainty of those predictions, which is crucial for guiding subsequent experiments and evaluating model reliability.

[0165] Model Definition: A Gaussian process is entirely defined by its mean function m(X) and covariance function k(X,X′). The prior mean function m(X) is set to 0, and the Matern 5 / 2 kernel function is chosen as the covariance function (kernel function) because it handles smooth changes in function values ​​well and is less sensitive to parameters. Its mathematical form is:

[0166] ;

[0167] The independent variable , and It is the input vector of any two sets of samples. It is the signal variance. It is a length scale parameter.

[0168] Model training: Maximum likelihood estimation (MLE) is used to optimize the hyperparameters of the kernel function. and noise level , where n is the total number of training samples.

[0169] The optimal hyperparameters are obtained by maximizing the logarithmic marginal likelihood function. The log-marginal likelihood function used is: ;

[0170] in For the kernel matrix, It is the identity matrix. , These are the input matrix and the output matrix, respectively. This is the transpose of the output matrix; the dimensions of each matrix are related to the number of samples. The hyperparameters of the kernel function are... denoted as the noise level of the kernel function, and n is the total number of training samples.

[0171] The trained GPR proxy model, for a new input mediator variable vector (New samples, also known as test points), their predicted changes in kinematic characteristics The conditional distribution of is a Gaussian distribution. The predicted mean of this Gaussian distribution is... (i.e., the most likely predicted value) and prediction variance (The uncertainty of the forecast) is given by the following key formula:

[0172] ;

[0173] ;

[0174] in, The kernel function calculates the similarity. Test point The n×1 kernel similarity vector between the sample and all training points X (i.e., the input vector corresponding to the sample). express Autocorrelation terms under the kernel function, Let y be the kernel matrix composed of training point samples, and y represent the change in kinematic features corresponding to the sample (i.e., the output value). These are terms pre-calculated during the training phase and can be viewed as a set of "weight" coefficients.

[0175] The essence of the prediction process: new samples The predicted value is a weighted average of the output values ​​y of all training samples, with the weights calculated by the kernel function k to represent the similarity. Decision. With The more similar the training samples, the greater their contribution to the prediction result. The kernel function plays a central role in defining "similarity" here.

[0176] Model validation: Leave-one-out cross-validation (LOOCV) was used to evaluate the predictive performance of the surrogate model. The root mean square error (RMSE) and coefficient of determination were calculated. () as an evaluation indicator.

[0177] , ;in, These represent the i-th true label value, the i-th model predicted output value, and the mean of the true labels, respectively; N is the total number of predicted values.

[0178] The model is required to achieve the following on the test set: >0.9, to ensure prediction accuracy.

[0179] This invention further integrates the simulation and model prediction processes described above, using a final global response surface model to directly express the correlation between tension band design parameters and kinematic characteristic changes. This global response surface model allows designers to input any set of tension band design parameters (pretension F, area A, direction...) This model can directly and quickly output the predicted change in motion performance (ΔM). The global response surface model consists of two pre-established sub-models connected in series:

[0180] The trained Gaussian process regression (GPR) surrogate model constitutes a mapping function from "mechanical mediator variables" to "changes in motion performance". ,Right now:

[0181] ; where the offset .

[0182] Combined with the previously established finite element simulation model from "design parameters" to "mechanical mediating variables" ;

[0183] A global response surface model can be obtained that directly predicts the final motion utility of the tension band from its design parameters:

[0184] ;

[0185] This composite model can predict any given design parameters (F, A, ...) of any tension band structure. Changes in kinematic characteristics caused by ) .

[0186] The following specific experimental case further illustrates the concrete implementation of the embodiments of the present invention. This experimental case is a utility prediction surrogate model for running shorts, in which the prediction is still based on the hip joint angular velocity change described above. For the goal.

[0187] The input feature (X) selects simulation results related to the gluteus maximus (R09) and biceps femoris (R16) as input:

[0188] ;

[0189] That is, each sample is an 8-dimensional vector.

[0190] The output (Y) is the change in hip joint angular velocity. (Unit: deg / s), calculated through video analysis.

[0191] The GPR model was trained using data corresponding to 20 sets of CCD parameter simulations (n=5 periods, 100 samples in total). The optimized Matern kernel function hyperparameters are: .

[0192] The model performance was evaluated using 5-fold cross-validation. The results show that the surrogate model predicts... The RMSE is 0.22 deg / s. =0.94, indicating that the model has extremely high prediction accuracy.

[0193] Example of performance evaluation application for tension band structure in new sportswear: Suppose we want to evaluate a new set of tension band parameters (F=13N, A=50cm²). =0.85) of utility.

[0194] Input this parameter into the established finite element simulation model (Or run a new simulation) to obtain the predicted value Xnew of the mechanical mediator variable vector;

[0195] Input Xnew into the trained GPR agent model ;

[0196] The model instantly outputs the prediction results: .

[0197] The prediction indicates that this set of parameters is expected to increase the average hip joint angular velocity by 3.4 deg / s, with a 95% confidence interval.

[0198] By combining parametric finite element simulation with a Gaussian process regression surrogate model, this invention successfully constructs an efficient and accurate design-performance prediction workflow. This method significantly reduces reliance on physical prototypes and biomechanical experiments, providing core technical support for the rapid digital design and utility evaluation of sportswear.

[0199] After establishing the mapping model and various evaluation indicators, the embodiments of the present invention can comprehensively evaluate the structural performance of sportswear.

[0200] The kinematic feature changes predicted based on the validated global response surface model ( Combining the principles of sports biomechanics, a multi-dimensional utility comprehensive scoring system is used to quantify the overall utility of sportswear.

[0201] Identification of positive and negative effects: First, based on the biomechanical characteristics of exercise, determine the changes in each characteristic ( The "benefit direction" of a characteristic is defined. The benefit direction defines how changes in this characteristic affect overall performance: changes consistent with the benefit direction are considered positive effects, while changes contrary to it are considered negative effects. See Table 6 below for details:

[0202] Table 6

[0203]

[0204] Targeted processing: Based on the benefit direction in the table above, the predicted original changes ( ) is given directionality, resulting in a directional change ( ), >0 indicates a performance improvement. <0 indicates a performance decrease.

[0205] Formula for calculating directional change:

[0206] ;

[0207] in: It is the absolute change of the k-th feature directly predicted by the model; It is the directional sign coefficient, determined based on the "benefit direction":

[0208] When the benefit direction is "+", =+1;

[0209] When the benefit direction is "-", =-1;

[0210] When the benefit direction is "depending on the type of sport", the expected direction needs to be clarified based on the specific type of sport before calculation.

[0211] Overall Utility Score Calculation: The overall utility score is a dimensionless numerical value. U > 0 indicates an improvement in overall performance, and U < 0 indicates a decrease in overall performance. The overall utility score (U) is calculated using the following formula:

[0212] ;

[0213] in, The directional change of the k-th feature; : The reference change of the k-th feature, where ; The weight coefficients of the k-th feature under a specific motion type (see table below) satisfy the following conditions: .

[0214] Exercise type weighting table ( As shown in Table 7 below:

[0215] Table 7

[0216]

[0217] Utility level classification:

[0218] Based on the scoring results, the following design suggestions are provided in Table 8:

[0219] Table 8

[0220]

[0221] Continuing with the example of evaluating running shorts, the case background is as follows: For a pair of running shorts (containing two tension bands)... [R09], [R16]) is evaluated, with the following parameters: ; .

[0222] Type of sport: Long-distance running

[0223] Step 1: Model Prediction and Utility Scoring

[0224] Input the parameters into the global response surface model to obtain the prediction results:

[0225] (Knee joint): +1.8 (Assuming that long-distance running aims to reduce the angle to increase stability, the benefit direction is -);

[0226] (Hip joint): +3.5deg / s (benefit direction +);

[0227] (Benefit direction-);

[0228] (Benefit direction-)

[0229] Calculate the directional change ( ):

[0230] ;

[0231] Select the long-distance running weights (w=[0.1, 0.3, 0.3, 0.3]) and calculate the overall score:

[0232] .

[0233] Results Analysis: The results were classified as "Good" (0.2 ≤ U ≤ 0.6). Analysis showed that the design significantly improved hip joint force efficiency and trunk stability (both reaching 70% of the reference threshold), but negative changes in knee joint range of motion and symmetry lowered the overall score, indicating a significant biomechanical compensatory effect in the current parameter configuration. Adjustment of the biceps femoris tension band is recommended. The pretension parameters are adjusted to effectively improve motion symmetry and knee joint coordination while maintaining core performance gains.

[0234] Compared with existing technologies, the present invention proposes a method for evaluating the performance of tension band structures in sportswear. By constructing a sportswear utility evaluation system based on the fusion of tension band structure and human posture characteristics, it forms a complete closed loop of "clothing design parameters – human movement performance – utility score – design optimization," which has the following significant advantages over existing technologies:

[0235] 1. A unified and standardized tension band design description system

[0236] This invention introduces a set of design parameters to describe the force, area, direction, angle, and region of action of each tension band, and establishes a standardized numbering library, enabling quantitative comparison of the structures of different sports and types of clothing within a unified framework. This method solves the problems of "experience-based design, inconsistent descriptions, and lack of data interoperability" in traditional sportswear, supporting product database construction and large-scale statistical analysis. It allows different design teams to share a unified design language, avoiding errors caused by inconsistent understanding of terminology or structures, and shortening cross-team collaboration cycles.

[0237] 2. Enables a label-free, natural evaluation process with wide applicability.

[0238] This invention utilizes a markerless posture recognition algorithm combined with an interactive correction mechanism to achieve accurate identification of key joints and movement features of athletes without relying on optical motion capture, inertial sensors, or human body marker attachments. The method is simple to operate, the evaluation process does not interfere with natural movement, and it is suitable for complex scenarios such as real training, outdoor testing, and competition analysis. It overcomes the limitations of laboratory reliance and high costs, reducing costs by more than 90% (requiring only a regular camera vs. a megapixel-level optical capture system), thus improving the practicality and scalability of the evaluation technology.

[0239] 3. Construct a structure-action fusion model to achieve interpretability analysis of clothing utility.

[0240] Using a technical approach of "parametric finite element simulation + Gaussian process regression surrogate model", a model was constructed based on the tension band design parameters (F, A, ...). This study establishes a quantitative and interpretable mapping relationship between tension band structures and biomechanical mediators (muscle stress, joint forces), and then to changes in kinematic characteristics. It clarifies the positive or negative contribution of tension band structures to specific movements, thereby enabling interpretable and quantifiable evaluation of the utility of functional clothing and providing a scientific basis for design optimization.

[0241] 4. Applicable to various action scenarios and professional sports, with strong versatility.

[0242] This invention possesses excellent versatility in its algorithm and modeling methods, applicable not only to the evaluation of basic movements (such as running, squatting, and pushing a cart), but also to competitive sports that demand higher levels of postural stability and explosive movement, such as bobsleigh, speed skating, and weightlifting. By appropriately configuring movement templates and analysis parameters, it can be flexibly extended to the adaptability evaluation of different sports and functional clothing.

[0243] 5. Low equipment requirements, controllable implementation costs, and easy operation.

[0244] Compared to traditional physiological assessment platforms or high-end motion capture systems, this invention only requires conventional shooting equipment (such as high-definition cameras, action cameras, or mobile phones) and structural parameter input (CAD drawings, image annotations, etc.). It does not require additional high-cost hardware equipment and is not subject to space constraints, providing a low-threshold and usable solution for small and medium-sized laboratories, corporate clothing testing, and grassroots sports units.

[0245] On the other hand, embodiments of the present invention provide a system for evaluating the structural performance of tension bands in sportswear, the system comprising:

[0246] The tension band quantification module is used to perform quantitative analysis on the sportswear to be evaluated, and to represent one or more tension bands in the sportswear to be evaluated using one or more corresponding structural parameter vectors.

[0247] The finite element simulation module is used to apply the structural parameter vectors of one or more tension bands to a parameterized human biomechanical three-dimensional model, and to conduct finite element simulation experiments through the parameterized human biomechanical three-dimensional model to collect the mechanical mediator variable vectors under the action of one or more tension bands.

[0248] The surrogate model module is used to make predictions in a pre-trained Gaussian process regression surrogate model using the mechanical mediating variable vector to obtain the changes in kinematic characteristics before and after wearing the sportswear to be evaluated.

[0249] The performance evaluation module is used to evaluate the overall performance of the sportswear to be evaluated based on the changes in the kinematic characteristics.

[0250] This invention provides a low-cost, highly adaptable, and explanatory functional sportswear utility evaluation system by constructing a mapping model that integrates tension band structure and human posture characteristics. In particular, it addresses the structure-motion coupling effect modeling problem of tension band clothing and has broad application potential in sports science research, product verification, individualized training, and smart wearable evaluation.

[0251] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0252] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for evaluating the structural performance of tension bands in sportswear, characterized in that, The method includes the following steps: A quantitative analysis is performed on the sportswear to be evaluated, and one or more tension bands in the sportswear to be evaluated are represented by one or more corresponding structural parameter vectors. The structural parameter vector includes at least the tension direction, effective area, and pretension of the tension band; Applying the structural parameter vectors of the one or more tension bands to a parameterized three-dimensional human biomechanical model includes the following steps: Each of the tension bands is modeled as a membrane unit in the parameterized three-dimensional human biomechanics model; The tension direction is reflected by the local material orientation of the membrane unit or the direction of the applied load; The effective working area is parametrically controlled by adjusting the thickness of the membrane unit and the covered grid area. The prestress is applied by setting the initial stress of the membrane unit in a predefined field; Finite element simulation experiments were conducted using the parametric human biomechanics 3D model to collect the mechanical mediating variable vectors under the action of one or more tension bands. The changes in kinematic characteristics before and after wearing the sportswear to be evaluated are obtained by using the aforementioned mechanical mediating variable vector in a pre-trained Gaussian process regression surrogate model. The overall performance of the sportswear to be evaluated is assessed based on the changes in the kinematic characteristics.

2. The method according to claim 1, characterized in that, The structural parameter vector also includes the tension band action area, and the mechanical mediating variable vector includes: the maximum stress of the target muscle, the maximum strain of the target muscle, the relevant joint reaction force, and the interfacial shear stress between the tension band and the skin in the tension band action area.

3. The method according to claim 2, characterized in that, The kinematic characteristics include: joint angles of at least one related joint, joint angular velocities of at least one related joint, trunk stability indices, and symmetry indices; wherein the at least one related joint is determined based on the tension band action area.

4. The method according to claim 1, characterized in that, The Gaussian process regression surrogate model uses the Matern5 / 2 kernel function as the covariance function. The pre-training optimizes the hyperparameters and noise level of the kernel function through maximum likelihood estimation, and evaluates the predictive performance of the surrogate model through leave-one-out cross-validation, so that the conditional distribution of the kinematic feature changes predicted by the Gaussian process regression surrogate model based on the mechanical mediating variable vector is a Gaussian distribution.

5. The method according to claim 4, characterized in that, The kernel function is: Among them, the independent variable , and It is the input vector of any two sets of samples. It is the signal variance. It is a length scale parameter.

6. The method according to claim 5, characterized in that, The maximum likelihood estimation obtains the optimal hyperparameters of the kernel function by maximizing the log-marginal likelihood function, which is: in For the kernel matrix, It is the identity matrix. , These are the input matrix and the output matrix, respectively. This is the transpose of the output matrix; the dimensions of each matrix are related to the number of samples. The hyperparameters of the kernel function are... denoted as the noise level of the kernel function, and n is the total number of training samples.

7. The method according to claim 6, characterized in that, The leave-one-out cross-validation method calculates the root mean square error (RMSE) and the coefficient of determination. As an evaluation indicator: ； in, Let be the i-th true label value, the i-th model predicted output value, and the mean of the true labels, respectively; N is the total number of predicted values.

8. The method according to claim 1, characterized in that, The overall performance of the sportswear under evaluation is assessed by calculating a comprehensive utility score U. in, This represents the directional change of the k-th kinematic feature; Let be the reference change for the k-th kinematic feature. Let k be the weight coefficient of the k-th kinematic feature under a specific motion type, satisfying d represents the total number of kinematic characteristics. It is the absolute change of the k-th kinematic feature directly predicted by the model. It is the coefficient of the benefit direction sign.

9. A system for evaluating the structural performance of tension bands in sportswear, characterized in that, The system includes: The tension band quantification module is used to perform quantitative analysis on the sportswear to be evaluated, and to represent one or more tension bands in the sportswear to be evaluated using one or more corresponding structural parameter vectors; the structural parameter vectors include at least the tension direction, effective area, and pretension of the tension band; The finite element simulation module is used to apply the structural parameter vectors of one or more tension bands to a parameterized three-dimensional human biomechanical model, modeling each tension band as a membrane element in the parameterized three-dimensional human biomechanical model; the tension direction is represented by the local material orientation or the direction of the applied load of the membrane element; the effective working area is parametrically controlled by adjusting the thickness of the membrane element and the mesh area it covers; the pretension is applied by setting the initial stress of the membrane element in a predefined field; and finite element simulation experiments are performed using the parameterized three-dimensional human biomechanical model to collect the mechanical mediating variable vectors under the action of one or more tension bands. The surrogate model module is used to make predictions in a pre-trained Gaussian process regression surrogate model using the mechanical mediating variable vector to obtain the changes in kinematic characteristics before and after wearing the sportswear to be evaluated. The performance evaluation module is used to evaluate the overall performance of the sportswear to be evaluated based on the changes in the kinematic characteristics.

Images